WARNING : this sample may be harmful for the hearing (*) Listen at your own risk !

Here is one of the strangest effect I've heard. For me, these samples are nearly impossible to distinguish with headphones, while the difference is more than obvious on speakers (ABX 8/8, not even in front of the speakers).

The difference is also obvious with headphones and the Dolby Headphones DSP (ABX 8/8), but not with the Bauer crossfader DSP (they sound the same).

I looked for the factor that could made the difference audible.
-Mp3 encoding : no
-Vorbis encoding : no
-Phase rotation : no
-peaks or dips in the frequency response : only with strong peaks. But the effect is not the same (not ABXed).
-reverberation : yes

The repetition of the same sample over and over seem to reveal an audible difference, but this difference disappear if the samples are played back with silences between the repetitions. I could not ABX it because I've got no ABX program that allows to play the samples looped. Anyway, the samples are different at the end, and should thus click differently when the playback loops.

(*) Listening to this sample, even at low volume, for some time also has a strange effect on my hearing. Everything sounds warbled after that. Including my tinnitus ! The effect lasts for a minute. I have no idea if this comes from a stress on the internal ear, which would be dangerous, or just a confusion in the brain.

An alien mind control sequence would exactly sound like that! Better don't play this, it could mess with your brai¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥¥bckspace¥¥¥¥¥¥¥¥
backspace key broken¿ Disregard all output.

All human being play back sample immediately. Tell friends.

My god, music only get's weirder these days...

Very odd. The two files are identical until a little past half way, at which point the B file "loops" (the data is the same as at the beginning of the file,) while the A file continues without any glitch. I wonder if it's noticeable if you're using speakers that don't share a subwoofer?

The very strange distorted hearing aftereffect is very likely a auditory analog to visual motion aftereffects. I'd never read anything about an audio version of that, but Googling "auditory aftereffect" returned quite a few hits, so I guess it's not an unknown phenomena.

So all these files appear to be are summed sinusoids at exponentially increasing frequencies - hitting every note on a keyboard basically. In such an environment the only difference that can originate between samples generated in the same way is in their phases; when B restarts, there is a transient pulse (because of the instantaneous phase modulation) followed by a different phase relationship compared to A.

So, when some kind of reverb is introduced - either artificially in the case of Dolby Headphone or physically in the case of speakers in a real room - the pulses involved are temporally smeared. If the pulses were previously inaudible, but were masked in a mostly temporal fashion, the reverb may extend them long enough for temporal masking to fail and for them to become more audible.

I'm also kinda wondering if static distortion effects (speaker THD, nonlinear mixing in the ear, etc) play a role in modulating around sinusoids with close-but-not-quite-matched phase relationships. That could yield the warbling Pio (and I) hear afterwards, and it could also make the instantaneous "sound" of the signal quite different between the two tones, especially around the pulses.

Did you use a crossfade to loop the sample ?
Experienced audio editors know that transitions between almost identical segments require a cross fade in order to be inaudible. The most important variables are the position and duration of the fade. If the fade is too short, it can result in clicks or pops. If the fade is too long, it can create phasing effects, up to the point where things can sound double. In many cases I've found edits that were inaudible on headphones to be audible on speakers and also the other way around.
If you don't have longer segments to use for your ABX test, it will be impossible to completely exclude the effects of the crossfade from your results.

Wow, this is interesting! Thanks for posting! Did you take this sample from RMAA? Maybe this is due to the longer impulse response of the low frequency speaker driver? It's so audible on my Tapco S-5s, not even worth doing an ABX. I'll play this over the most expensive speakers I can find at work on Monday and report back.


They do, at least on my equipment.

I think this is some form of auditory adaptation, something like the Zwicker tone.

Chris

Yes, I noticed the effect accidentally, while looping the RMAA frequency response test tone. So I made sample B looping the first half of sample A.



My speakers have no subwoofer and the effect is very audible.



A spectrogram shows clearly what's going on :







No, the transition in the middle of sample B is sharp. The signal jumps from a value to the other.

In the ABX, I listen to each sample once.



I don't think so, because I can hear the transition in the middle of sample B if I put down the headphones, while I can't hear it wearing them.

You're right, same for me. Now that is scary! Might have implications on audio coding, i.e. artifacts audible over speakers which are inaudible over headphones.

Chris

There is an explanation for this:
Some frequencies just before and after the splice might not be in phase (which happens very often, not to say most of the time). When you add reverb, the signal before the splice will be continued, while the other signal (not in phase) will be mixed with it. This inevitably leads to spectral gain changes.
This effect happens with (long) crossfades as well. Some frequencies are amplified, others reduced.
There is no cure-all solution for making inaudible crossfades. That's the reason that DAWs aimed at audio editing offer so many crossfade options.

Interesting theory. It seems plausible.
In this case, the "bleep" of the B file would behave like the echo. With long reverberations, it would sound longer, and if we setup a reverb not strictly decreasing, it would follow the amplitude variation of the echo. For example with a reverberation first decreasing, then increasing, then decresing again.

Some experiment to test Pio's finding and Kees' theory: I applied Audition's "Medium Auditorium" reverb preset to EtrangeB.flac. When played over headphones, does it sound like over speakers?

It does. Actually other presets (but not all of them) worked as well. So it seems you guys are right.

Quite shocking how room response can alter sound...

Click to view attachment

Chris

I remember such a behaviour also. On one of the first tracks of "Massives Attack - Mezzanine" i heard a mid-low frequency plop on my speakers i barely could hear even after knowing where it is on my headphones.
Sadly i must have borrowed this cd to someone a while now and can't listen for it atm.

Now i got it. In Angel at ~4.40 it was but i have to admit i hear it easily on headphones now... wonder why.
One thing, with headphones it sounds more like a small click due to a ripping error but it isn't. On speakers it sounds completely different with much more low-frequemcy energy.